*Manuscript Replication Instructions

*Manuscript text replication on ISIS support

*Although there may be self-censoring of responses, a total of 91% of IDP camp respondents indicated some form of ISIS affiliation to include having worked for ISIS (21%), fought for ISIS (17%), having family members who worked for ISIS (55%) and having family members who were put on trial for their role in ISIS (64%).

sum workedisis spouseisis famisis foughtisis spousefoughtisis famfoughtisis famtrialisis

tab isisties
*(note: isisties coded 1 if any of the previous ISIS related variables were answered positively, 0 if otherwise)

*Figure 1

cibar revtrustgov, over1(mosul) 
graph save g1, replace
cibar revsupportisis, over1(mosul) 
graph save g2, replace
graph combine g1.gph g2.gph
*(Note: additional formatting required)

*Results – Manuscript discussion in text

*Consistent with H2, two sample t-tests indicate that the average punishment in the revenge treatment is significantly greater than in the security treatment (t =6.11, p<0.0000). The average punishment ranges between a short and long-term prison sentence in the revenge treatment (mean = 2.58, sd = 1.14) compared to short-term prison in the security threat treatment (mean = 1.98, sd =1.23).

ttest killrevthreat, by(treatment) unequal
sum killrevthreat if treatment==1
sum killrevthreat if treatment==2

*In both subsamples, there is greater support for punishment in the grievance treatment than in the security threat treatment, but the treatment effect is much stronger among Mosul civilians (t =15.07, p<0.0000) than IDP camp respondents (t = 2.39, p<0.0096).

ttest killrevthreat if mosul==1, by(treatment) unequal
ttest killrevthreat if mosul==0, by(treatment) unequal

*In particular, Mosul civilians’ remarkably strong preferences for amnesty in the security treatment (87%) indicates solid support for informal retributive violence against suspected insurgents, constitutes a clear disregard for international laws and norms, and does not bode well for peace and security.

tab killrevthreat if mosul==1 & treatment==2

*Figure 2

ciplot killrevthreat if mosul==1, by(treatment) 
graph save g1, replace
ciplot killrevthreat if mosul==0, by(treatment) 
graph save g2, replace
graph combine g2.gph g1.gph
*(Note: additional formatting required)

twoway (histogram killrevthreat if treatment==1, by(mosul) discrete percent) (histogram killrevthreat if treatment==2, by(mosul) discrete percent fcolor(none) lcolor(black))
*(Note: additional formatting required as well as combining of sub-graphs to create Figure 2)

*Table 1

ologit killrevthreat treatment, robust
ologit killrevthreat treatment mosul, robust
ologit killrevthreat treatment revisisnoquit revviolencejust revwillofpeople revblameciv isisvictim female age education income professional laborer unemployed moved, robust

*Footnote 8

*Only 7% of respondents strongly disagree that violence is ever justifiable, and the item is not significant predictor of punishment preferences in Table 1.

tab revviolencejust

*Results – Judicial Killing in text discussion

*When comparing basic treatment effects across this survey item, we find that subjects in the surrender treatment are less likely to see capital punishment as a fair outcome than those in the capture treatment (two-sample t-test = 10.48, p<0.0000). However, when interacting treatments with ISIS affiliation via Mosul vs. IDP camp comparisons, Figure 3 shows that the treatment effect only applies to the subsample of Mosul civilians (t =10.12, p<0.0000). There is no significant treatment effect for respondents in IDP camps outside Mosul (t = 0.05, p<0.4782).

ttest revrightcapsur, by(treatment) unequal
ttest revrightcapsur if mosul==1, by(treatment) unequal
ttest revrightcapsur if mosul==0, by(treatment) unequal

*Figure 3 indicates that less than 10% of IDP camp respondents strongly believed that the judge made the right decision handing down the death penalty.

tab revrightcapsur if mosul==0

*Figure 3

ciplot revrightcapsur if mosul==1, by(treatment) 
graph save g1, replace
ciplot revrightcapsur if mosul==0, by(treatment) 
graph save g2, replace
graph combine g2.gph g1.gph
*(Note: additional formatting required)

twoway (histogram revrightcapsur if treatment==1, by(mosul) discrete percent) (histogram revrightcapsur if treatment==2, by(mosul) discrete percent fcolor(none) lcolor(black))
*(Note: additional formatting required as well as combining of sub-graphs to create Figure 3)

*Results - In text discussion

*Many IDP camp respondents indicate that they preferred either a long-term sentence (58%), a short-term prison sentence (24%) or even amnesty (15%) in lieu of the death penalty in both treatments. As an alternative to capital punishment, most Mosul civilians indicated they would accept a life sentence (79%) or long-term incarceration (13%) in the capture treatment. In the surrender treatment, a slim majority (54%) still support a life sentence, but nearly one-third (32%) would support long-term imprisonment as an alternative to death.

tab fadhildecision if mosul==0
tab fadhildecision if mosul==1


*Table 2

ologit revrightcapsur treatment, robust
ologit revrightcapsur treatment mosul, robust
ologit revrightcapsur treatment revisisnoquit revbiasedisis revdeathpenjust revblameciv isisvictim female age education income professional laborer unemployed moved, robust

*Footnote 11

*Few Iraqis (13%) strongly oppose the death penalty, while 80% agree that the death penalty can be justified underscoring widespread support for retributive violence in the judicial system. 

tab revdeathpenjust


*Appendix Table 1. Summary of Variables

sum killrevthreat revrightcapsur revisisnoquit revviolencejust revwillofpeople revbiasedisis revdeathpenjust revblameciv isisvictim female age education income professional laborer unemployed student moved


*Online Appendix Material

*Index of Support for Execution Response Items

*Index construction

factor faircapsur rightcapsur justcapsur revrevengecapsur revfadhildecision
alpha faircapsur rightcapsur justcapsur revrevengecapsur revfadhildecision
*(Note: indices already created in dataset using the gen function)

graph bar faircapsur rightcapsur justcapsur revrevengecapsur revfadhildecision if mosul==0 
graph bar faircapsur rightcapsur justcapsur revrevengecapsur revfadhildecision if mosul==1

*ISIS Surrender/Stop Fighting Index

factor amnesty pardonpows amnestylib neglib
alpha amnesty pardonpows amnestylib neglib
factor fairtrial notorture nodeathpenalty lessprison protectisis protectiraqi financialaid
alpha fairtrial notorture nodeathpenalty lessprison protectisis protectiraqi financialaid
*(Note: indices already created in dataset using the gen function)

*Victimization Index

graph bar punishedis familypunishedis injuredis familyinjuredis familykilledis imprisonedis fleehomeis hometakenis womenabusedis if mosul==1
graph bar punishedis familypunishedis injuredis familyinjuredis familykilledis imprisonedis fleehomeis hometakenis womenabusedis if mosul==0

factor punishedis familypunishedis injuredis familyinjuredis familykilledis imprisonedis fleehomeis hometakenis womenabusedis
*(Note: index generated already in dataset using the predict command)

*Treatment Effects Estimation

*Balance tests

ksmirnov female if mosul==1, by(treatment)
ksmirnov female if mosul==0, by(treatment)
ksmirnov age if mosul==1, by(treatment)
ksmirnov age if mosul==0, by(treatment)
ksmirnov education if mosul==1, by(treatment)
ksmirnov education if mosul==0, by(treatment)
ksmirnov income if mosul==1, by(treatment)
ksmirnov income if mosul==0, by(treatment)
ksmirnov professional if mosul==1, by(treatment)
ksmirnov professional if mosul==0, by(treatment)
ksmirnov laborer if mosul==1, by(treatment)
ksmirnov laborer if mosul==0, by(treatment)
ksmirnov unemployed if mosul==1, by(treatment)
ksmirnov unemployed if mosul==0, by(treatment)
ksmirnov moved if mosul==1, by(treatment)
ksmirnov moved if mosul==0, by(treatment)

*Treatment Effects

*Manuscript Table 1 Model 3

teffects ra (killrevthreat mosul revisisnoquit revviolencejust revwillofpeople revblameciv isisvictim female age education income professional laborer unemployed moved) (treatment)

teffects ipwra (killrevthreat mosul revisisnoquit revviolencejust revwillofpeople revblameciv isisvictim female age education income professional laborer unemployed moved) (treatment female age education income professional laborer unemployed moved)

teffects ipwra (killrevthreat) (treatment mosul female age education income professional laborer unemployed moved)

*Manuscript Table 2 Model 3

teffects ra (revalphacapsur mosul revisisnoquit revbiasedisis revdeathpenjust revblameciv isisvictim female age education income professional laborer unemployed moved) (treatment)

teffects ipwra (revalphacapsur mosul revisisnoquit revbiasedisis revdeathpenjust revblameciv isisvictim female age education income professional laborer unemployed moved) (treatment female age education income professional laborer unemployed moved)

teffects psmatch (revalphacapsur) (treatment mosul female age education income professional laborer unemployed moved)

*Mosul vs. Camp Sampling Locations Comparisons

logit mosul revisisnoquit revviolencejust revwillofpeople revblameciv isisvictim female age education income professional laborer unemployed moved, robust

logit mosul revisisnoquit revbiasedisis revdeathpenjust revblameciv isisvictim female age education income professional laborer unemployed moved, robust

*Mosul vs. Camps on Likelihood of ISIS Surrender/Stop Fighting Index

cibar revisisnoquit, over1(mosul)

twoway  (histogram revisisnoquit if mosul==0, percent) (histogram revisisnoquit if mosul==1, percent fcolor(none) lcolor(black)) 
*(Note: additional formatting required)

*Mosul vs. Camps on Blame Attribution

cibar revblameciv, over1(mosul)

twoway  (histogram revblameciv if mosul==0, discrete percent) (histogram revblameciv if mosul==1, discrete percent fcolor(none) lcolor(black))

*Mosul vs. Camps on other significant IVs in Manuscript Tables 1.

cibar revwillofpeople, over1(mosul)

twoway  (histogram revwillofpeople if mosul==0, discrete percent) (histogram revwillofpeople if mosul==1, discrete percent fcolor(none) lcolor(black))

*Mosul vs. Camps on other significant IVs in Manuscript Tables 2.

cibar revbiasedisis , over1(mosul)

twoway  (histogram revbiasedisis if mosul==0, discrete percent) (histogram revbiasedisis if mosul==1,discrete percent fcolor(none) lcolor(black))

cibar revdeathpenjust , over1(mosul)

twoway  (histogram revdeathpenjust if mosul==0, discrete percent) (histogram revdeathpenjust if mosul==1,discrete percent fcolor(none) lcolor(black))

*Manuscript Table 1 Robustness Checks

cibar killrevthreat, over1(treatment) over2(mosul)
*(Note: additional formatting required)

*Punishment for Extrajudicial Killing (OLS, Ordered Logit, Multinomial Logit Regression)

reg killrevthreat treatment, robust
ologit killrevthreat treatment, robust
mlogit killrevthreat treatment, robust

*Punishment for Extrajudicial Killing (OLS, Ordered Logit, Multinomial Logit Regression, Controlling for Mosul vs. Camps)

reg killrevthreat treatment mosul, robust
ologit killrevthreat treatment mosul, robust
mlogit killrevthreat treatment mosul, robust

*Punishment for Extrajudicial Killing (OLS, Ordered Logit, Multinomial Logit Regression, Extended Controls)

reg killrevthreat treatment revisisnoquit revviolencejust revwillofpeople revblameciv isisvictim female age education income professional laborer unemployed moved, robust
ologit killrevthreat treatment revisisnoquit revviolencejust revwillofpeople revblameciv isisvictim female age education income professional laborer unemployed moved, robust
mlogit killrevthreat treatment revisisnoquit revviolencejust revwillofpeople revblameciv isisvictim female age education income professional laborer unemployed moved, robust


*Punishment for Extrajudicial Killing (Logit regression)

logit drevthreat treatment, robust
logit drevthreat treatment mosul, robust
logit drevthreat treatment revisisnoquit revviolencejust revwillofpeople revblameciv isisvictim female age education income professional laborer unemployed moved, robust

*Manuscript Table 2 Robustness Checks

reg revalphacapsur treatment, robust
reg revrightcapsur treatment, robust

reg revalphacapsur treatment mosul, robust
reg revrightcapsur treatment mosul, robust

reg revalphacapsur treatment revisisnoquit revbiasedisis revdeathpenjust revblameciv isisvictim female age education income professional laborer unemployed moved, robust
reg revrightcapsur treatment revisisnoquit revbiasedisis revdeathpenjust revblameciv isisvictim female age education income professional laborer unemployed moved, robust

log close



